Manioc agriculture and sedentism in Amazonia: the Upper Xingu example.

Heckenberger, Michael J.

Agricultural productivity and Amazonian settlement

The nature of Pre-Columbian agricultural systems in Amazonia has stimulated considerable debate, specifically: can one or another cultigen - maize or manioc - provide a stable agricultural base for sedentism and population growth (e.g. Cameiro 1961; 1986; Gross 1975; Meggers 1996; Roosevelt 1980)? Certain ecological factors are generally seen to limit production and intensification of those subsistence resources that can support sedentary or densely distributed populations. Low agricultural productivity, characteristic of many Amazonian soils, and the generally low density and patchy distribution of terrestrial game are commonly cited as limiting factors (Gross 1975; 1983; Johnson 1982; Meggers 1954; 1996; Ross 1978; Sponsel 1989). It has become accepted that the highly restricted varzea regions, primarily the floodplain settings of the major 'white-water' rivers (the Amazon and its Andean-derived tributaries), did not impose these environmental constraints on demographic or economic growth due to their fertile soils and higher concentrations of rich aquatic resources (e.g. Brochado 1984; 1989; Cameiro 1986; 1995; Denevan 1996; Lathrap 1968; 1970; 1987; Lathrap et al. 1985; Meggers 1996; Moran 1993; Roosevelt 1980; 1989; 1994).

In areas away from the varzea, the broadly-defined terra firme, the model of environmental limitation still has wide currency: '[u]nder extensive management systems, the increased labour time cost of subsistence production provides a strong incentive for groups to keep their settlement density low and move about frequently (520 year intervals)' (Gross 1983: 429). Many scholars, in fact, feel the incentive was so great as effectively to have prevented the development of large, permanent villages in upland areas characterized by low-fertility soils (e.g. Brochado 1989; Brochado & Lathrap 1982; Gross 1975; Meggets 1996; Roosevelt 1980; 1991; 1994). Certainly local ecology differentially constrains cultural development in Amazonia, but we adequately understand neither the parameters of ecological variability nor of cultural adjustments to it. The limits imposed on human populations by Amazonian nutrient cycles turn out to be substantially higher than analysis of contemporary societies might lead us to believe (e.g. Balee 1989; 1995; Beckerman 1979; Denevan 1992; Dufour 1994; 1995); manioc agriculture, even on low-fertility soils (utisols/oxisols), when combined with ample aquatic resources can provide a stable economic foundation for densely settled, fixed populations (see, especially, Cameiro 1957; 1961; 1986; 1995; Denevan 1996; Lathrap et al. 1985).

Heated debate continues over the subsistence base of Amazonian societies, particularly with respect to:

* the staple crops of the densely populated varzea societies [manioc, maize, or some similar seed-crop) and

* whether large, sedentary societies could develop elsewhere in the region; if so, did this also depend on staple foods other than manioc?

Debate is deadlocked because reconstruction of prehistoric diet remains largely speculative; most models are derived not from detailed analysis of an actual prehistoric pattern, but instead projected from ethnographic and human ecological patterns of contemporary groups. Those few studies which use archaeological data are often preliminary and depend almost exclusively on one or a few well-documented examples. Expanding interpretations from specific archaeological examples to broad regional patterns therefore depends on gross, and often unwarranted, generalizations regarding contemporary populations.

In the present discussion, ethnographic and archaeological evidence are reviewed which demonstrate:

* large, densely populated, and fully sedentary populations did exist in one non-varzea setting (the Upper Xingu) prehistorically;

* these populations relied primarily on bitter manioc and fish, as do their ethnographically known descendants; and

* increased productivity of these staple foods using an existing technology, and not technological innovation, provided the economic foundation for populous, sedentary social formations.

The Upper Xingu (Mate Grosso, Brazil), one of the southernmost extensions of the hyleian forests of Amazonia, is presently home to diverse cultural groups, speaking Arawak, Carib, Tupian, and other languages. Most Upper Xingu communities are part of a distinctive Xinguano regional cultural pattern, established in the region over a thousand years ago. Since initial recorded contacts in the late 1800s, the Xinguanos have been extensively studied ethnographically. Most important, for the present discussion, are Robert Carneiro's (e.g. 1957; 1961; 1983) studies of Xinguano (Kuikuru) settlement patterns and subsistence, particularly manioc agriculture; these studies were the first to challenge the environmental limitations theory by empirically refuting the assumption of low agricultural productivity for Amazonian soils. Carneire (1983: 103) concludes that in the past:

The Kuikuru and their neighbors probably had bigger villages, stronger chiefs, mobilized labor on a larger scale, and perhaps even had social classes. If manioc cultivation did not create this culture, it at least provided the economic foundation on which it could be reared.

Carneiro's ethnographic interpretations have met with mixed feelings (see Descola 1996; Gross 1975; Roosevelt 1980), but there is no a priori ecological reason to doubt them (Beckerman 1979: 535). Preliminary archaeological investigations conducted in the 1950s by Carneire and Gertrude Dole (1961-2) documented earthworks in the Upper Xingu clearly related to large, settled villages; due to misinterpretation of these as natural features by several professional archaeologists who later worked in the region (Knoip 1969; Simoes 1967), the significance of the sites has seldom been considered in critiques of Carneiro's reconstructions. New archaeological evidence, based on recent investigations by the author (summarized in Heckenberger 1996), supports and elaborates Carneiro's conclusions regarding village size and the productivity of Xinguano subsistence patterns, especially manioc agriculture.

Villages and village life in diachronic perspective

The cultural sequence in the Upper Xingu, as presently known, extends from c. AD 900 until the present. Within this sequence, cultural continuity is demonstrable based on conservatism within three fundamental, and archaeologically visible, aspects of Xinguano culture:

* ceramic technology;

* village spatial organization; and

* settlement location within the Upper Xingu basin.(1)

The earliest occupations (c. AD 900-1500) apparently relate to colonization of the Upper Xingu by Arawakan groups from the west; they represent the ancestral foundation of contemporary Xinguano culture. Over the past 500 or more years, various distinctive groups moved into the basin; many came to share this basic Xinguano cultural pattern (Heckenberger 1996; 1997). Beyond demonstration of cultural continuity, these resilient features of regional cultural patterns provide the basis to reconstruct changes in regional demography, economic organization, and the scale and productivity of agriculture. Continuity in local ceramic industries and land-use patterns demonstrate that the basic economic orientation, based primarily on terra firme agriculture and exploitation of diverse aquatic resources, characterized communities throughout the sequence. Dramatic changes have occurred during the c. 1000-year Xinguano occupation, most obvious when we consider the size and regional distribution of contemporaneous settlements.

Contemporary villages are formed by a ring of houses around a large circular plaza. Plazas range from roughly 100 to 250 m in diameter; typically between 50 and 350 people occupy a village. Plaza villages are situated in selected areas which provide access to diverse ecological settings, notably the upland forests and major waterways (and associated bottomland); these favoured locations become focal points for long-term permanent occupations [ILLUSTRATION FOR FIGURE 1 OMITTED]. Thus, the ideal Xinguano settlement pattern involves permanent occupation and long-term transformation of these 'choice' areas, in an ideal of sedentism that permeates every aspect of village life. Unlike many Amazonian societies, particularly those which emphasize mobility (seasonal or permanent) and hunting, Xinguanos prefer a large, well-maintained and permanent ('beautiful') village, depend on the continued exploitation of diverse habitats around these villages, and delight in the camaraderie and sociability provided by stable and well-developed patterns of trade and interaction. Settled village life not only reflects itself in village spatial organization, subsistence economics and socio-political relations, but also permeates ritual life, with pronounced intra- and inter-village ceremonialism, and the personalities of villages.

The picture provided from archaeology demonstrates that in the past Xinguanos were also sedentary manioc farmers and fisher-people; and it documents that their settled tropical-forest life-way was not held in demographic (or evolutionary) stasis by ecological limitations. Manioc agriculture, in combination with the rich aquatic resources of the Upper Xingu, provided the economic foundation for substantially larger village populations and regional population aggregates, as Carneiro predicted. Archaeological investigations conducted by the author at various sites in the Upper Xingu basin were concentrated in the territory (some 625 sq. km) currently occupied by one contemporary Kuikuru (Carib-speaking Xinguano) village. The Kuikuru - or Carib groups ancestral to the contemporary Kuikuru - have occupied villages in one or another part of this territory for at least the past c. 500 years. Considerable fieldwork was conducted at two sites in Kuikuru territory, Nokugu (MT-FX-06) and Kuhikugu (MT-FX-11)(2) [ILLUSTRATION FOR FIGURE 2 OMITTED].

The obvious difference between contemporary villages and prehistoric sites is scale: prehistoric villages are immense by comparison. Contemporary Xinguano villages are commonly situated within or adjacent to ancient settlements. Some 2.5 km west of the present-day Kuikuru village - the largest historically-known Xinguano village in area V, roughly 275 m in diameter, and in population some 320 people - lies the prehistoric site of Nokugu. At Nokugu, a vast area of over 40 ha and extending over a kilometre along the Anahuku (Buriti) River is covered by archaeological remains [ILLUSTRATION FOR FIGURE 3 OMITTED]. At first glance, one would think that the archaeological distributions resulted from the type of clustered occupations typical of contemporary settlement patterns. Upon thorough exploration, one sees that the entire site area of Nokugu is covered by ceramic sherds and by various artificial earthen constructions. Systematic mapping of these features, made possible by opening parallel linear corridors (5 m wide) through a wide array of vegetation, reveals that the earthworks articulate into an elaborate structural plan. A similar pattern can be reconstructed for the even larger site of Kuhikugu (some 50 ha in area], where the Kuikuru lived from the mid 1800s until 1961(3) [ILLUSTRATION FOR FIGURE 4 OMITTED], and at several other sites in the Kuikuru study area (MT-FX-13, 17, 18, and - perhaps - others).

Mapped in their full extent, the substantial earthworks, apparently constructed more or less contemporaneously across the region in the 14th century, provide a ready plan for archaeologists of late prehistoric settlement organization. Their intentional construction according to an integrated architectural plan is unmistakable. More structurally elaborated than in contemporary villages, this plan nonetheless reflects the same underlying concentric model of spatial organization; domestic areas gravitate toward a central plaza(s) with spoke-like radial causeways. Late prehistoric communities, unlike their ethnographic descendants, constructed huge barriers around their settlements. At Nokugu and Kuhikugu, and other sites in the Upper Xingu,(4) the most prominent earthworks are large, semicircular ditches at the outer margin of the domestic occupation area and, at Nokugu, also within the village area. These ditches reach a depth of 3-4 m, including the ridge of excavation overburden heaped on the inside berm; the village peripheral ditches often extend over 2 km. Linear ridges, roughly 0.5-2 m high, at the edges of central plazas and intra-village causeways also are prominent features in the integrated architectural plan. A primary function of this plan was almost certainly defence, although these features undoubtedly had important aesthetic, symbolic and, perhaps, economic functions.

The layout of prehistoric villages or the earthworks themselves is of less interest here than what they say about village size and permanence. The scale of the earthworks shows that the communities which built them had every intention of staying put; whether or not the villages had been abandoned and reoccupied during the centuries preceding earthwork construction (c. AD 900-1000 to AD 1400). Peripheral ditches define the boundaries of these ancient villages since virtually all cultural remains (e.g. ceramics) are confined within the ditches. The distribution of domestic ceramics throughout areas within the peripheral ditch(es) demonstrates that these roughly define the boundaries of residential occupation, although occupation was apparently denser in areas closer to the central plazas. The lack of anthropogenic 'black earth' (terra preta) or subsurface ceramics in most areas closer to the peripheral ditch(es) indicates that these areas were occupied late in time, suggesting population nucleation concomitant with ditch construction.

Although we cannot estimate village populations precisely, the physical scale of the sites and their associated earthworks indicates substantially larger populations than in historically known villages. The contemporary Kuikuru village, about 0.06 sq. km or 6.0 ha, is about one-eighth the size of the average of Nokugu and Kuhikugu, 0.45 sq.km; the village had a population of about 320 in 1994. This figure multiplied by eight would give a population estimate of about 2500 people. The within-village Kuikuru population density of about 53 persons per hectare multiplied by the Nokugu-Kuhikugu average of 45 ha gives a similar figure, 2385 people (see Agostinho 1993:277 for even higher estimates).

These estimates of village population based strictly on village area are conjectural. Nonetheless, we can conclude that late prehistoric villages, more than 10 times the area of most historically-known villages, had higher populations; particularly considering that the plaza constitutes well over half of the entire village area in the Kuikuru village (roughly 5 ha) and considerably less in late prehistoric villages. The substantial earthworks, and the aboveground structures (e.g. palisades, fences, bridges, traps) which likely accompanied them, would have demanded a large labour force, likely much larger than could be mounted today. I would guess that village population at sites like Nokugu, Kuhikugu and others likely exceeded 1000-1500 people, and may have ranged considerably higher.

Xinguano subsistence economy

Clearly village and regional populations were substantially higher in late prehistoric times. Likewise, it is clear that these populations were fully sedentary. The question remains, how did these large population aggregates support themselves? Contemporary populations combine manioc farming in terrafirme areas with aquatic food foraging, and there is compelling reason to believe that prehistoric populations depended on the same staple foods. Contemporary Xinguano subsistence patterns, like most ethnographic Amazonian populations, show a great diversity in the foraged and produced resources; nevertheless, bitter manioc and fish together constitute 90-95% of the diet (Carneiro 1994).

Subsistence practices

Cultivated plants make up some 85% of the Xinguano diet, the vast majority from some 46 varieties of bitter manioc (Manihot esculenta sp.) (Carneiro 1983; 1994: 207). Numerous other domesticates, including maize (Zea mays), sweet potatoes (Ipomoea bataras), hot peppers (Capsicum sp.), pineapple (Ananas comosus), squash (Cucerbita maxima), peanuts (Arachis hypogaea), banana (Musa paradisiaca) and recently introduced non-native cultigens, are grown. Among non-food domesticates, Xinguanos cultivate gourd (Lagenaria siceraria), cotton (Gossypium barbadense), tobacco (Nicotiana sp.), and uructi (Bixa orellana). Palm fruits, including buriti (Mauritia flexuosa) and macatiba (Acrocomia sclerocarpa), as well as piqui (Caryocar brasiliense) and other fruits, e.g. goiriba (Psidium guajava), caju (Anacardium occidentale) and mangaba (Hancornia speciosa), are also consumed (Carneiro 1978). Some, notably piqui and macauba, are considered 'semi-domesticates' since they occur in areas of past habitation sites, but do not generally occur naturally. Salt is manufactured from the ashes of water hyacinth (aguape) and from burned tree termite nests.

Buriti palm, which grows in low-ground wet areas, is of primary importance not only for its fruits; its leaves and stalks are used for myriad purposes (e.g. cordage, mats, seats, skirts, internal house walls, occasionally house thatch). Sape grass (Imperata sp.), a colonizer of disturbed ground, is also of singular importance as house thatch. The house frame is constructed using select forest trees (Carneiro 1978). Innumerable other wild plants are used for industrial purposes, medicines and other purposes.

Second to agriculture in terms of subsistence activities and dietary importance is fishing. Fish, including over 50 exploited species, provide about 10-15% of the diet, the primary source of animal protein (see Camelto 1957; 1994: 207). Traditional Xinguano technology, which includes various traps, weirs, nets, poison, bow and arrow, and lances, can exploit virtually all the diverse fishing areas - small streams and rivers, major rivers, and standing bodies of water ranging from small 'mud-holes' to expansive and deep lakes (Basso 1973: 37-9; Carneiro 1957: 126-31). In recent times, hook-and-line fishing, a western introduction, has become important.

Other forms of animal protein include: wasp larvae, sauva ants, several species of grasshoppers (sometimes taken in large quantities), turtles (Podocnesis sp.) and their eggs, one type of monkey (Cebus sp.) and several species of birds, including jacu (guan) and curassow (Cracidae), and one or two smaller species. Monkeys and birds are especially sought during times when dietary restrictions preclude the consumption of fish (see Carneiro 1957: 116-25).

By and large, Xinguanos (especially Carib and Arawak groups) have strong restrictions against the consumption of 'red' meat from the abundant terrestrial animals (including tapir, peccary, deer, sloth, paca, capyvara, coati, agouti, armadillo, tortoise) (Basso 1973; Carneiro 1957; Carvalho 1951; Gregor 1977). Restrictions also apply to some aquatic animals, notably cayman and very large fish. Dietary restrictions (taboos) vary slightly from village to village depending on location and, especially, ethnic group.

Manioc processing and ceramic continuity

Manioc is processed and cooked through a sophisticated process that is, by and large, unique to the Upper Xingu. In brief summary, manioc processing proceeds through the following general steps:

* tubers are collected from gardens and brought to the village;

* tubers are peeled (traditionally using a shell);

* peeled tubers are grated using a wooden grater (inagi);

* grated pulp is rinsed through a mat strainer (tuafi) over a large vessel separating the pulp into a fine fraction (which passes through the strainer) and a heavy fraction (which does not pass through the strainer and is removed and dried as a clump called timbuku);

* the water from the straining is removed and boiled for several hours to remove toxins and produce a thick beverage called kuigiko.

* the fine fraction is removed in sections and dried;

* the dried sections are ground into flour (kuiginu); timbuku is also sometimes ground into flour but only if the kuiginu runs out.

(For detailed descriptions of manioc processing see Carneiro 1983: 96-9; Dole 1978.)

Manioc processing, technically sophisticated and standardized, is functionally linked to a very specific set of utensils and cookware. Contemporary Xinguano communities recognize three primary ceramic forms which, although historically manufactured almost entirely by the Arawakan-speaking Waura, are found in every household throughout the Upper Xingu. The forms include:

* large to medium, flaring-rim vessels (averaging 45-70 cm in mouth diameter), which are used to process and cook manioc [ILLUSTRATION FOR FIGURE 5 OMITTED];

* medium to small, everted- or direct-rim cooking vessels (averaging 20-45 cm in mouth diameter), which are used primarily to cook fish; and

* flat or shallow griddles (averaging 35-55 cm in mouth diameter).

Native communities likewise recognize these as the primary ceramic forms. The Kuikuru (Carib-speaking Xinguano) classify these three principal forms as ahukugu, atangi and alato; the Waura likewise recognize these primary forms, called kamalupi, makula and heshe, respectively. Size and rim form are the primary characteristics the Kuikuru use to distinguish between ah ukugu and atangi forms. These forms have specific functions in contemporary villages, namely processing and cooking of manioc (ahukugu and alato) and fish (atangi), although any ceramic vessel can serve a variety of ancillary functions; of note, ahukugu forms are also used to cook piqui fruits during harvesting season (late October through early December), and a variant was traditionally used to carry and store water (see Galvao 1953: 49).

These 'core' elements of the Xinguano ceramic industry are virtually identical in prehistoric and in contemporary ceramic assemblages; the primary ethnographic forms correspond exactly to the principal forms in prehistoric assemblages, notably including:

* the large to medium cooking pots, identical to ahukugu forms (designated here as Type 1A) [ILLUSTRATION FOR FIGURE 6 OMITTED];

* the medium to small (atangi-like) cooking pots, with everted (Type 1B) and direct (Type 1C) rims; and

* manioc griddles.

Of all vessels identified from the prehistoric ceramic assemblage (n = 688 forms), Type I vessels dominate with 584 forms, 85%. Of the Type 1 forms, the 363 Type 1A vessels constitute 62% (53% of the total inventory of typed vessels; with 209 definite forms from Mt-Fx-06 and 154 forms from Mt-Fx-11), and the 206 Type lB and 1C vessels 35% (30% of the total inventory; with 136 forms from Mt-Fx-06 and 70 forms from Mt-Fx-11). Griddles constitute about 6% of the total assemblage, and roughly 8% of the total number of typed vessels correspond to forms other than ah ukugu, atangi or alato forms. Continuity with contemporary ceramic industries can be demonstrated also in ceramic manufacture: paddle/anvil (slab/clump) manufacturing technique, the preponderance of cauixi temper (sometimes mixed with other inclusions, including burned tree bark (caraipe), grog, charcoal and grit], techniques of surface treatment (exterior red slip, interior black pigment), burnishing and smoothing, and decoration (rim incising, punctuation and rim adorno applique). Small quartz burnishing stones and clay manufacture 'clumps' recovered in excavated prehistoric deposits likewise resemble modern examples and provide further evidence of ceramic continuity.

We can assume that the economic functions related to these ceramics correspond to similar functions in prehistoric communities and, like form, manufacture and decoration, ceramic use-wear patterns on Type 1A ceramic vessels correlate well between the prehistoric and the contemporary ceramic assemblages. The two most informative wear patterns are:

* significant erosion on the inside rim and vessel interior, likely due to the processing or cooking of bitter manioc notable for its acidic toxins; and,

* wide notches worn into the rim lips of large ahukugu forms resulting from the cross-bars placed over the pots to process manioc.

Functional continuity, obvious in the manioc-processing ceramics (ahukugu or Type 1A forms and griddles), can also be suggested for the fish-processing ceramics (atangi or Type lB and 1C forms). Although the introduction of aluminium and plastic vessels has resulted in the near-total replacement of ceramics for water-carrying, storage and some food-processing activities, cooking of manioc and fish continues to be conducted almost exclusively in ceramic vessels due to their functional superiority.

Sufficient archaeological data demonstrates that bitter manioc was the staple food of prehistoric Xinguanos by three lines of evidence. There is demonstrable continuity in ceramic ware associated with manioc processing. Except for seasonal piqui and infrequent sweet potato processing, plant products are generally not processed or cooked in ceramic vessels; although maize fiat cakes are rarely cooked on griddles, for example, maize is roasted over an open fire and ground into flour in large wooden mortars. Secondly, as widely recognized (see Roosevelt 1980:79-93 for a summary), the generally low fertility of unaltered Amazonian upland soils (e.g. oxisols/utisols) makes them unsuitable for the cultivation of many crops (notably maize and other seed crops). Finally, the use of terra firme (non-inundated) forest settings near prehistoric sites, provides the means to evaluate prehistoric manioc agriculture.

The limits of Xinguano subsistence productivity

Xinguano communities, like most Amerindian peoples, exist in intricate anthropogenic landscapes formed in the natural environment by the long-term occupation of specific ecological settings (Posey & Balee 1989). Settlements are, and have been in the past, located at the interface of terra firme forest (providing flat, dry forest land for manioc gardens) with rivers, lakes and even small streams (providing areas for fishing, fresh water and transportation). As localized areas experience concentrated long-term occupations, the already mosaic regional ecology becomes even more patchy; specific locales - chosen because they meet select hydrological, topographic and vegetational criteria and are therefore naturally good habitation locations - become even more attractive for human habitation. Pronounced anthropogenic alternation of the environment concentrates both natural and symbolic resources in these special places (Heckenberger 1998). Over time, intimate familiarity with these concentrated resources induces continually greater commitment to place. This pronounced commitment to and alteration of specific places relates to a general Xinguano ethos of sedentism characteristic of both past and present communities (described above), but prehistoric villages were much larger, more permanent and elaborated, and transformed local ecology far more than contemporary villages. This raises important historical questions, most notably for the present discussion: what are the productive limits of a subsistence technology based on manioc farming and fishing?

Late prehistoric settlements being substantially larger than contemporary villages -and more numerous - necessitates that overall production must have been considerably higher. This does not necessitate, however, that individual families produced more than present-day Xinguanos, that production was more communally based, or that the primary subsistence technology was radically different. Economic intensification involved increased productivity of the primary food sources, manioc and fish, rather than innovations in technology or the introduction of exotic cultigens. Increased agricultural productivity could have been achieved either by more extensive (more gardens) or more intensive farming (greater productivity per unit area). Likewise, aquatic foraging could have been practised more extensively or more intensively through management systems (e.g. turtle pens). Overall increases in productivity quite likely involved all of the above to some degree, as well as greater reliance on resources other than manioc or aquatic fauna.

Carneiro (e.g. 1957; 1961; 1983) has shown how agricultural productivity within a domestic mode of production similar to today could be increased far beyond historically-known levels - more extensive agriculture. He also provides ethnographic documentation that manioc is, in fact, produced, processed and stored in great quantities during the dry season (Carneiro 1983). Contemporary manioc gardens (kwigi anda) are opened in areas of primary or secondary forests, used for 2-5 years, after which the plot usually stands fallow for about 10-30 years. Manioc gardens are well tended during their use, and often substantial barricades are constructed to keep peccaries out. Fallow periods do not represent complete abandonment, but alteration with a long-term cycle. Gardens are often managed through continued reburning to produce sape grass for house thatch, or as groves of piqui trees, planted while the plot is still producing manioc. Gardens not managed after abandonment quickly return to scrub forest (after 10-30 years) and high forest (50-70) years; it is these secondary forest stands, and not primary forest, which are usually cleared for new gardens, only rarely being opened in 'primary' or 'virgin' forest (itsuni) (Carneiro 1983). So, Upper Xingu communities do not abandon their gardens at all, although the crops (domesticated or not) planted or managed in them change over time in a long-term pattern of cyclical rotation.

Long-term alteration of local vegetation, clearly documented in aerial photographs, demonstrates that agricultural areas around prehistoric villages were far more extensive than those associated with contemporary villages [ILLUSTRATION FOR FIGURES 7 & 8 OMITTED]. It is also clear that these areas were used much more intensively than today, as well. Forest alteration related to historic occupations has a low impact on local forest cover and more closely corresponds to the general Amazonian (ethnographic) model of extensive swidden agriculture - discrete clearings in the forest that, as Carneiro (1983) noted, return to high forest in less than a century. In contrast, the prehistoric agricultural areas, like the prehistoric village areas themselves, have still not returned to high forest after centuries of disuse and show a patchy or mosaic regrowth. Although difficult to evaluate precisely, prehistoric manioc farmers undoubtedly used their garden areas far more intensively than the pattern of short-cropping followed by long fallows or abandonment characteristic today (see also Denevan 1992).

In a subsistence economy based on manioc and fish, what was the nutritional status of the Upper Xingu diet? As suggested by Dufour (1995: 160), manioc 'may not be quite so poor a source of protein and minerals as we have assumed, but as is true of other staple foods its nutritional value is a function of processing'. In the Upper Xingu, as in most areas of Amazonia, the processing techniques that remove the toxic properties of bitter manioc provide a wide range of edible substances each with distinctive nutrient characteristics (Carneire 1983; Dole 1978). Fine-grained manioc flour (kuiginho) produced by contemporary Xinguanos is more highly processed even than that produced by most other Amazonian groups; it undoubtedly has a nutritional status rather different from the unprocessed raw, cooked or fermented sweet manioc from which most studies of the nutritional status of manioc are derived (Dufour 1994; 1995]. Nonetheless, manioc products do not in and of themselves provide a balanced diet and we must consider the ability to increase harvesting of primary subsistence resources other tha manioc. The contemporary Xinguano diet, based on manioc and fish, maintains good health among communities numbering in the hundreds (Dufour 1994: 164-5; Fagundes-Neto et al. 1981), but could it be maintained among village populations numbering in the low thousands given a technology and dietary profile like that of contemporary Xinguanos?

With respect to the cultivation of secondary crops other than manioc, another aspect of agricultural systems should be addressed: the house garden (see Denevan 1992; Lathrap 1997). House-midden 'black earth' deposits within villages are only minimally used by contemporary Xinguanos for house gardens. These deposits, due to their high fertility, support a wide variety of subsidiary crops, many of which do not grow well in unaltered terrafirme soils (e.g. maize). In the Kuikuru village, some backyard house middens are only minimally cultivated, others are mono-cropped, but most are planted with varied food and other plants (e.g. gourds and tobacco). In prehistoric settlements, rich 'black earth' soils, which are far more extensive than in contemporary villages and also include large contiguous blocks associated with the road and plaza marginal mounds (formed in part by repeated dumping of domestic refuse), may have been cultivated even more systematically and intensively by families or larger social groups.

In contemporary villages, the primary protein source is fish, supplemented by turtles, turtle eggs, insects (ants, grasshoppers, wasp larvae), birds and diverse plant products (notably palm, piqui and other fruits). The relative quantities of Type lB and 1C ceramic vessels (fish-cooking pots) demonstrates that fish were a primary food source prehistorically as they are today. Many aquatic resources could be harvested at levels far surpassing contemporary levels using a technology essentially identical to those of the present-day Xinguanos. Communal fishing with weirs and traps, provides the means to harvest substantial quantities of fish. Primary traps, utu and itaka, are placed in weirs using a stick or pole frame with palm leaf thatch placed between the wooden braces below the water-line. The largest Kuikuru weir, cutting off the Anahuku River at Ahanitahugu, was a tall (3-6 m) community-built weir several hundred metres in length and containing over 40 conical itaka traps. During high-water fish runs, hundreds of kilogrammes of fish could be harvested in a single day; swimmers would also occasionally drive fish downstream into the traps. Smaller, squat weirs with one or a few itaka or utu traps were constructed by individuals or a related group of men for private use. Weirs are also placed across standing bodies of water to facilitate fishing with poison (inte) or with a plunge basket-trap (kundu). Several hundred kilogrammes of fish were captured for one ceremonial payment from the shallow waters (less than i metre) at Hialugihiti using the kundu dunk traps. Prominent men also organize fishing expeditions by a large group of related men, typically netting hundreds of kilogrammes of fish and normally carried out today using commercial nets. Little attention has been paid to issues of increasing productivity of aquatic resources in native Amazonian economies, especially in broadly-defined terrafirme areas (cf. Garson 1980; Limp & Reidhead 1979); the Upper Xingu case demonstrates the vast potential of these resources (Carneire 1986).

Fishing productivity varies seasonally, but alternative nutrient sources can also be used or produced at much higher levels than is ethnographically known (see Beckerman 1979; Carneire 1957). Native cultivated plants other than manioc - maize, piqui, sweet potato, peanuts, peppers, among others - can also be collected in large quantities; like manioc, many are storable for long periods. Piqui is harvested in great quantity and stored as pulp under water. Carneire (pers. comm., 1994), for instance, witnessed over 11,000 piqui fruits collected by the Kuikuru in preparation for an egitse ceremony the following year. Other seasonally available foods include several varieties of palm fruits (available in different seasons), mangaba, turtles (sometimes kept in pens or tethered with a string through the carapace) and turtle eggs, as well as insects/larvae. In sum, we can conclude that in the Upper Xingu, processed bitter manioc and fish, supplemented with diverse wild plant and animal foods, prehistorically provided a secure nutritional base for large, sedentary regional populations.

Discussion

For better or worse, prominent debates in Amazonian archaeology, or more broadly, in the study of Amazonian cultural development, revolve around the potential of one or another agricultural economy for sedentism and population growth (Carneire 1995). As it is acknowledged that large, sedentary societies existed along the Amazon (the varzea) in late prehistoric and early historic times (and likely long before), recent debate has focused on differences, particularly the differential productivity of soils and aquatic resources, between this area and the rest of Amazonia. Comparisons are often flawed, however, insofar as they ignore prehistory and context and tend uncritically to compare ancient occupations of the varzea with contemporary upland occupations. Yet there are no comprehensive early ethnohistorical or archaeological studies which document that regional settlement patterns were roughly similar to present-day patterns, from anywhere in the Amazonian uplands.

Even in those few areas where well-documented sequences of cultural development can be reconstructed, the dietary base is uncertain. Significant archaeological evidence exists: manioc-based systems can be inferred from analysis of lithic artefacts, graterboard chips, and ceramic griddles (cf. DeBoer 1975); Roosevelt's (1980, 1991, 1994) arguments for maize or other seed-crop agricultural systems in the varzea are based on analysis of macro-botanical remains and on human bone chemistry. But precise reconstructions of prehistoric diet are hampered by severe problems of taphonomy, recovery and sampling. In the Upper Xingu, for example, manioc, which is largely processed outside houses and unburned refuse heaped on backyard trash middens, would be archaeologically detectable through microbotanical or soil chemical analyses (rarely employed in the tropical lowlands), but would likely be largely absent from macrobotanical samples; conversely, maize (a minor crop) would be obvious in macrobotanical analyses since ample carbonized remains are likely produced when maize is roasted in indoor firehearths.

Carneiro's discussions of Xinguano cultivation practices provide one of the most detailed treatments of agricultural potential in Amazonia (e.g. 1961; 1983). He showed how manioc production in the Upper Xingu, using a technology essentially identical to that of contemporary communities, could sustain large (up to 2000 according to him) sedentary populations, but it remained to be demonstrated that it did. Archaeological evidence, now available and summarized here, generally substantiates Carneiro's conclusions; it refutes general models which minimalize the potential of upland areas for intensive agriculture (e.g. Gross 1975; Meggers 1996; Ross 1978) or that suggest intensive agricultural systems and dense, sedentary social formations in upland settings were narrowly restricted to areas of high-fertility soils (e.g. Roosevelt 1991; 1994a). It also supports the general observation, often overlooked in discussions of protein limitations in Amazonia, that where aquatic resources abound, fishing and other forms of aquatic foraging generally take economic precedence over terrestrial hunting (Beckerman 1994; Carneiro 1986; Lathrap et al. 1985).

In the Upper Xingu, demonstrated continuity in cultural and material life provides the means to evaluate long-term trends through controlled historical comparison (rather than broad general, and often uncritical, correlations or analogies). What is clear from such comparisons is that local populations were not held in demographic stasis by any broadly defined ecological determinant (e.g. soil infertility or protein limitation) or technological deficiency. In fact, although it seems reasonable that such an ecological imperative may have characterized some regions of Amazonia prehistorically, this remains adequately to be demonstrated archaeologically for any area of Amazonia. Likewise, the present findings do not refute models that propose fundamental transformations in subsistence economies as the primary catalyst of prehistoric cultural change (e.g. a purported shift in some varzea economies from root-crop to seed-crop agricultural systems; Roosevelt 1980; 1991; 1994). But they do cast serious doubt on the generality of these models. Cultural change in the Upper Xingu, at any rate, is not closely correlated with such transformations, but instead relates to the consequences of other social, political or ideological factors. The Xinguano subsistence economy remained essentially the same through periods of profound change in village and regional demography and socio-political relations; in fact, in historical perspective, Xinguano techno-economics, particularly settlement patterns and the manioc farming/fishing subsistence economy, turn out to be surprisingly flexible, in terms of scale and productivity, but extremely resistant to overall transformational change.

Research in the Upper Xingu underscores the diversity of resource management and settlement strategies among pre-columbian (and contemporary) Amazonian populations. A serious problem with archaeological reconstructions (and those from ethnology and human ecology) in the region stems from uncritical and generally unwarranted generalizations regarding upland ecologies and human interaction with them. The terra firme regions represent a tremendous range of geophysical, hydrological and biotic diversity, so variable regionally that it is questionable whether terra firme (or varzea for that matter) actually refers to any definable set of ecological characteristics (Denevan 1984; Moran 1991; 1995). Perhaps, as Carneiro (1994: 64) suggests, the Upper Xingu represents 'something of a "Halfway house" between the varzea habitats of the lower Amazon and the terra firme habitats of the interfluves', but there is no reason to see the Upper Xingu as an extraordinary, singular case - i.e. the conditions for intensified production of manioc and aquatic resource capture (and hence the potential for large, sedentary populations) apply to other upland areas as well. Therefore, the distinction between varzea and terra firme is a gross oversimplification, particularly insofar as it is seen directly to reflect differential potential for agricultural intensification, population growth or other related factors typically viewed as key determinants of cultural evolution.

In short, the empirical evidence necessary for archaeological reconstruction of generalized cultural patterns, such as the notion of a 'tropical forest culture', whether it defines riverine patterns (e.g. Lathrap 1977), those of upland areas (e.g. Roosevelt 1980; 1991), or both (e.g. Meggers 1996], simply does not exist. Regional anthropological debate is therefore better served by moving beyond overly general cultural models (i.e. prime mover or prime inhibitor models) and oversimplified generalizations to the development of in-depth studies of specific regions and regional social systems, in as much cultural and historical detail as possible, providing the basis for more penetrating and illuminating comparisons in the future.

Acknowledgements. Primary funding for the Upper Xingu research was generously provided by the National Science Foundation, Washington (grant no. DBS-9214806) and the Social Science Research Council, New York. My special thanks to Afukaka Kuikuru and his family, the Kuikuru community, Aritana Yawalipiti, Robert Carneiro, Bruna Franchetto, Sandra Wellington, James Richardson III and James Petersen.

1 The Upper Xingu basin can be generally characterized as a peneplain surrounded by topographically higher areas and is largely confined to the broad area (some 40,000 sq. km) where the principal headwater tributaries converge to form the Xingu River proper.

2 Site designations correspond to the nationwide Brazilian site registration system based on state, in this case Mato Grosso (MT), and region (here referring to headwaters (formadores) of the Xingu River (FX).

3 The Kuikuru moved to the village of Ahanitahagu, adjacent to Nokugu, to be included in the confines of the Xingu Indigenous Park which was established in 1961. They moved to their present location adjacent to Lake Ipatse, several kilometres east of Nokugu, in 1971.

4 Ditches and other works described here are known from at least 15 sites in the Upper Xingu (Heckenberger 1996: 76).

References

AGOSTINHO DA SILVA, P. 1993. Testemunhos da ocupacao ore-Xinguana na bacia dos formadores do Xingu, in V.P. Coelho (ed.), Karl van den Steinen: Um Seculo de Antropologia no Xingu: 223-87. Sao Paulo: Editera da Universidade de Sao Paulo.

BALEE, W. 1989. The culture of Amazonian forests. in Posey & Balee (ed.): 1-21.

1995. Historical ecology in Amazonia, in Spansel (ed.): 97-110.

BASSO, E.B. 1973. The Kalapalo Indians of central Brazil. New York (NY): Holt, Rinehart & Winston.

BECKERMAN, S. 1979. The abundance of protein in Amazonia: a reply to Gross, American Anthropologist 81: 533-60.

1994. Hunting and fishing in Amazonia: hold the answers, what are the questions?, in Roosevelt (ed.): 177-200.

BROCHADO, J.P. 1984. An ecological model of the spread of pottery and agriculture into eastern South America. PhD. dissertation, Department of Anthropology, University of Illinois at Urbana-Champaign.

1989. A Expansao dos Tupie da Ceramica da Tradicao Policromica Amazonica, Dedalo 27: 65-82.

BROCHADO, J.P. & D.W. LATHRAP. 1982. Amazonia. Unpublished MS. Department of Anthropology, University of Illinois at Urbana-Champaign.

CARNEIRO, R.L. 1957. Subsistence and social structure: an ecological study of the Kuikuru Indians. Ph.D dissertation, University of Michigan. Ann Arbor (MI): University Microfilms.

1961. Slash-and-burn cultivation among the Kuikuru and its implications for cultural development in the Amazon Basin, in J. Wilbert (ed.), The evolution of horticultural systems in native South America: causes and consequences, a symposium: 47-67. Caracas: Sociedad de Ciencias Naturales La Salle. Antropologica, Supplement Publication 2.

1978. The knowledge and use of rain forest trees by the Kuikuru Indians of central Brazil, in Ford 1978: 201-16.

1983. The cultivation of manioc among the Kuikuru Indians of the Upper Xingu, in Hames & Vickers 1983: 65-111.

1986. The ecological basis of Amazonian chiefdams. Unpublished ms.

1994. Kuikuru, in J. Wilbert (ed.), Encyclopedia of world cultures 7: South America, 206-9. Boston (MA): G.K. Hall.

1995. The history of ecological interpretations of Amazonia: does Roosevelt have it right?, in Sponsel (ed.): 45-65.

CARVALHO, J.C.M. 1951. Relacoes entre os Indies do Xingu e a fauna regional. Rio de Janeiro: Museu Nacional. Publicacoes Avulsas.

DEBOER, W. 1975. The archaeological evidence for manioc cultivation: a cautionary note, American Antiquity 40(4): 419-33.

DENEVAN, W. 1984. Ecological heterogeneity and horizontal zonation of agriculture in the Amazon floodplain, in M. Schmink & C. Wood, Floodplain expansion in Amazonia: 311-36. Gainsville (FL): University of Florida Press.

1992. Stone vs metal axes: the ambiguity of shifting cultivation in prehistoric Amazonia, Journal of the Steward Anthropological Society 20(1-2): 153-65.

1996. A bluff model of riverine settlement in prehistoric Amazonia, Annals of the Association of American Geographers 86(4): 654-81.

DESCOLA, P. 1994. In the society of nature: a native ecology in Amazonia. Cambridge: Cambridge University Press.

DOLE, G.E. 1961-2. A preliminary consideration of the prehistory of the Upper Xingu basin, Bevista do Museu Paulista. (NS)13: 399-423.

1978. The rise of manioc among the Kuikuru: some interpretations, in Ford 1978: 217-47.

DUFOUR, D.L. 1994. Diet and nutritional status of Amazonian peoples, in A. Roosevelt (ed.), Amazonian Indians from Prehistory to the Present: Anthropological Perspectives, 151-76. Tucson (AZ): University of Arizona Press.

1995. A closer look at the nutritional implications of bitter cassava use, in L.E. Spansel (ed.), Indigenous peoples and the future of Amazonia: an ecological anthropology of an endangered world: 149-65. Tucson (AZ): University of Arizona Press.

FAGUNDES NETO, U., et al. 1981. Observation of the Alto Xingu Indians (central Brazil) with special reference to nutritional evaluation in children. American Journal of Clinical Nutrition 34: 2229-35.

FORD, R.I. (ed.). 1978. The nature and status of ethnobotany. Ann Arbor (MI): Museum of Anthropology, University of Michigan. Anthropological Papers 67.

GALVAO, E. 1953. Cultura e sistema de parentesco das tribes do Alto Xingu, Boletim do Museu National (Antropologia) (NS) 14: 1-56.

GARSON, A.G. 1980. Comment upon the economic potential of fish utilization in riverine environments and potential archaeological biases, American Antiquity 45(3): 562-7.

GREGOR, T. 1977. Mehinaku: the drama of daily life in a Brazilian In dian village. Chicago (IL): University of Chicago Press.

GROSS, D. 1975. Protein capture and cultural development in the Amazon Basin, American Anthropologist 77(3): 526-49.

1983. Village movement in relation to resources in Amazonia, in Hames & Vickers 1983: 429-49.

HAMES, R. & W. VICKERS (ed.). 1983. Adaptive responses of native Amazonians. New York (NY): Academic Press.

HECKENBERGER, M.J. 1996. War and peace in the shadow of empire: sociopolitical change in the Upper Xingu of southeastern Amazonia, ca. AD 1400-2000. Ph.D dissertation, University of Pittsburgh. Ann Arbor (MI): University Microfilms International.

1997. A periferia meridional da Amazonia: cultura e desenvolvimento cultural na fronteira entre Brasfi e a bacia amazemica. Paper presented at the 9th annual meeting of the Sociedade de Arqueologia Brasileira, Rio de Janeiro, September 1997.

1998. Revisionism in Amazonian ecological anthropology: the ancient cultural landscapes of the southern Amazon. Unpublished manuscript.

JOHNSON, A. 1982. Reductionism in cultural ecology: the Amazon case, Current Anthropology 23 (4): 413-28.

KNEIP, L.M. 1969. Relaterio sabre as valetas do Porque Nacional do Xingu. Estado de Mate Crosse. Rio de Janeiro. UFRJ-Museu Nacional.

LATHRAP, D.W. 1968. The 'hunting' economies of the tropical forest zone of South America: an attempt at historical perspective, in R.B. Lee & I Devore (ed.). Man the hunter: 23-9. Chicago (IL): Aldine.

1970. The Upper Amazon. New York (NY]: Praeger.

1977.Our father the cayman, our mother the gourd: Spinden revisited, or a unitary model for the emergence of agriculture in the New World, in C.A. Reed (ed.), Origins of agriculture: 713-51. The Hague: Mouton.

LATHRAP, D.W., A. GEBHART-SAYER & A.M. MESTER. 1985. The roots of the Shipibo art style: three waves on Imariacocha or there were 'Incas' before the Incas, Journal of Latin American Lore 11(1): 31-119.

LIMP, W.F. & V.A. REIDHEAD. 1979. An economic evaluation of the potential of fish utilization in riverine environments, American Antiquity 44(1): 70-8.

MEGGERS, B.J. 1954. Environmental limitation on the development of culture, American Anthropologist 56: 801-24.

1996. Amazonia: man and culture in a counterfeit paradise, Revised edition. Washington (DC): Smithsonian Institution Press.

MORAN, E.F. 1991. O estudo da adaptacao humana em ecossistemas amazonicas, in W.A. Neves (ed.), Origens, adaptacoes e diversidade biologica do homen nativo da AmazOnia: 161-78. Belem: Museum Paraense Emilio Goeldi.

1993. Through Amazon eyes: the human ecology of Amazonian populations. Iowa City (ID): University of Iowa Press.

1995. Disaggregating Amazonia: a strategy for understanding biological and cultural diversity, in Spansel 1995: 72-95.

POSEY, D.A. & W. BALEE (ed.). 1989. Resource management in Amazonia; folk and indigenous strategies. New York (NY]: New York Botanical Society. Advances in Economic Botany 7.

ROOSEVELT, A.C. 1980. Parmona: prehistoric maize and manioc subsistence along the Amazon and Orinoco. New York (NY): Academic Press.

1989. Resource management in Amazonia before the Conquest: beyond ethnographic projection, in Posey & Balee (ed.): 30-62.

1991. Moundbuilders of the Amazon: geophysical archaeology on Marajo Island, Brazil. New York (NY): Academic Press.

1994a. Amazonian anthropology: strategy for a new synthesis, in Roosevelt 1994b: 1-29.

1994b. (Ed.), Amazonian Indians .from prehistory to the present: anthropological perspectives. Tucson (AZ): University of Arizona Press.

ROSS, E.B. 1978. Food taboos, diet, and hunting strategy: the adaptation to animals in Amazon cultural ecology, Current Anthropology 19(1}: 1-36.

SIMOES. M.F. 1967. Consideracoes preliminares sabre a arqueologia do Alto Xingu, in Programa Nacional de Pesquisas Arqueologicas: resultados preliminares do primeire one. 1965-1966. Belem, Museu Paraense Emilio Goeldi, vol. 6: 129-144.

SPONSEL, L.E. 1989. Farming and foraging: a necessary complementarity in Amazonia?, in S. Kent (ed.), Farmers as hunters: the implications of sedentism: 37-45. Cambridge: Cambridge University Press.

(Ed.). 1995. Indigenous peoples and the future of Amazonia: an ecological anthropology of an endangered world. Tucson (AZ): University of Arizona Press.